White Papers

Proxy ARP is supported for both unicast and broadcast ARP requests. Control packets, other than ARP requests destined for
the VLT peers that reach the undesired and incorrect VLT node, are dropped if the ICL link is down. Further processing is not
done on these control packets. The VLT node does not perform any action if it receives gratuitous ARP requests for the VLT
peer IP address. Proxy ARP is also supported on secondary VLANs. When the ICL link or peer is down, and the ARP request for
a private VLAN IP address reaches the wrong peer, the wrong peer responds to the ARP request with the peer MAC address.
The IP address of the VLT node VLAN interface is synchronized with the VLT peer over ICL when the VLT peers are
up. Whenever you add or delete an IP address, this updated information is synchronized with the VLT peer. IP address
synchronization occurs regardless of the VLAN administrative state. IP address addition and deletion serve as the trigger events
for synchronization. When a VLAN state is down, the VLT peer might perform a proxy ARP operation for the IP addresses of
that VLAN interface.
VLT nodes start performing Proxy ARP when the ICL link goes down. When the VLT peer comes up, proxy ARP stops for the
peer VLT IP addresses. When the peer node is rebooted, the IP address synchronized with the peer is not flushed. Peer down
events cause the proxy ARP to commence.
When a VLT node detects peer up, it does not perform proxy ARP for the peer IP addresses. IP address synchronization occurs
again between the VLT peers.
Proxy ARP is enabled only if you enable peer routing on both the VLT peers. If you disable peer routing by using the no
peer-routingcommand in VLT DOMAIN node, a notification is sent to the VLT peer to disable the proxy ARP. If you disable
peer routing when ICL link is down, a notification is not sent to the VLT peer and in such a case, the VLT peer does not disable
the proxy ARP operation.
When you remove the VLT domain on one of the VLT nodes, the peer routing configuration removal is notified to the peer. In
this case, the VLT peer node disables the proxy ARP. When you remove the ICL link on one of the VLT nodes using the no
peer-link command, the ICL down event is triggered on the other VLT node, which in turn starts the proxy ARP application.
The VLT node, where the ICL link is deleted, flushes the peer IP addresses and does not perform proxy ARP for the additional
LAG hashed ARP requests.
VLT Nodes as Rendezvous Points for Multicast
Resiliency
You can configure VLT peer nodes as rendezvous points (RPs) in a Protocol Independent Multicast (PIM) domain.
PIM uses a VLT node as the RP to distribute multicast traffic to a multicast group. Messages to join the multicast group (Join
messages) and data are sent towards the RP, so that receivers can discover who the senders are and begin receiving traffic
destined for the multicast group.
To enable an explicit multicast routing table synchronization method for VLT nodes, you can configure VLT nodes as RPs.
Multicast routing needs to identify the incoming interface for each route. The PIM running on both VLT peers enables both the
peers to obtain traffic from the same incoming interface.
You can configure a VLT node to be an RP using the ip pim rp-address command in Global Configuration mode. When you
configure a VLT node as an RP, the (*, G) routes that are synchronized from the VLT peers are ignored and not downloaded
to the device. For the (S, G) routes that are synchronized from the VLT peer, after the RP starts receiving multicast traffic via
these routes, these (S, G) routes are considered valid and are downloaded to the device. Only (S, G) routes are used to forward
the multicast traffic from the source to the receiver.
You can configure VLT nodes, which function as RP, as Multicast source discovery protocol (MSDP) peers in different domains.
However, you cannot configure the VLT peers as MSDP peers in the same VLT domain. In such instances, the VLT peer does
not support the RP functionality.
If the same source or RP can be accessed over both a VLT and a non-VLT VLAN, configure better metrics for the VLT VLANs.
Otherwise, it is possible that one VLT node chooses a non-VLT VLAN (if the path through the VLT VLAN was not available
when the route was learned) and another VLT node selects a VLT VLAN. Such a scenario can cause duplication of packets.
ECMP is not supported when you configure VLT nodes as RPs.
Backup RP is not supported if the VLT peer that functions as the RP is statically configured. With static RP configuration, if
the RP reboots, it can handle new clients only after it comes back online. Until the RP returns to the active state, the VLT
peer forwards the packets for the already logged-in clients. To enable the VLT peer node to retain the synchronized multicast
routes or synchronized multicast outgoing interface (OIF) maps after a peer node failure, use the timeout value that you
configured using the multicast peer-routing timeout value command. You can configure an optimal time for a VLT
node to retain synced multicast routes or synced multicast outgoing interface (OIF), after a VLT peer node failure, using the
multicast peer-routing-timeout command in VLT DOMAIN mode. Using the bootstrap router (BSR) mechanism, you
can configure both the VLT nodes in a VLT domain as the candidate RP for the same group range. When an RP fails, the VLT
peer automatically takes over the role of the RP. This phenomenon enables resiliency by the PIM BSR protocol.
916
Virtual Link Trunking (VLT)